Air Pollution Monitoring via Wireless Sensor Networks: The Investigation and Correction of the Aging Behavior of Electrochemical Gaseous Pollutant Sensors

Round 1

Reviewer 1 Report

This manuscript examines the correlation between the measurements from low-cost gas sensors and those from the reference instrument during the aging of the sensor and proposes a corrective formula to alleviate the impact of the aging on the accuracy of measurements. It can be published after addressing following questions.

1.      Section 3.1, the reference data is based on point B which is different from the location where the low-cost sensor measurement was conducted (point A). Is it possible that the local variability of the gas concentration can affect the data comparation?

2.      Line 254, what is the “official instrumentation installed”? Is it the reference instrument in point B?

3.      Table 2, the dataset in Jul-21 is significantly different from the datasets measured in other time periods especially for the NO₂. How to explain this difference?

4.      Equation 6, how to calculate the corresponding sensitivity at a specific temperature S_current?

Author Response

We would like to thank the reviewer for the time he/she spent to study our manuscript. His/Her comments and suggestions were carefully considered so that the answers given and the actions undertaken address all issues stated. 

 

1. Section 3.1, the reference data is based on point B which is different from the location where the low-cost sensor measurement was conducted (point A). Is it possible that the local variability of the gas concentration can affect the data comparation?

Respond to Comment 1:

The distance between A and B is denoted. The validity of the measurements’ comparison is supported by relevant research works mentioned. Specifically, the following text was added to clarify this issue:

“The distance between these two locations is 900m. Both locations share the same urban conditions. Actually, the spatial coverage of a monitoring site represents the quantification of the variability of concentrations of a specific pollutant around the site [71], [72]. while the assessment of representativeness aims at the delimitation of areas of the concentration field with similar characteristics at specific locations as well the spatial surrogate data (similar emissions sources and land-cover characteristics) [73], [74], [75].”

 

2. Line 254, what is the “official instrumentation installed”? Is it the reference instrument in point B?

Respond to Comment 2:

Indeed “official instrumentation installed” is the reference instrument in point B {added to Figure 2}. To clarify this point the following text was added:

 “…the official instrumentation installed (i.e. the measuring stations of PERPA – Point B) [70]…”

 

3. Table 2, the dataset in Jul-21 is significantly different from the datasets measured in other time periods especially for the NO₂. How to explain this difference?

Respond to Comment 3:

The reviewer has spotted the issue that activated the Authors to study the field of aging. An important issue that is not discussed sufficiently in the literature is the impact of the season on the sensors accuracy. Such a field is still under investigation in the literature. The very low NO₂ R² value during July remains a field of study but it is not affecting current work since as can be seen it is not deteriorating or improving after applying the aging compensation equations.  It is important to notice that even after the compensation and correction of the factors the R² is not becoming significantly higher for this case. The temperature limitations of the sensor vendors are also an issue that affects this performance.

In order to clarify this issue the following text was added:

“It becomes evident from Table 2 that NO₂ values during July show low cross correlation. As it will be discussed later on the impact of the aging compensation equations on the NO₂ is positive as it improves the corresponding R² despite the fact that it remains at low values.  Furthermore, in Greece, during the summertime, the environmental temperature push the sensors at their functional limits. Specifically, according to the manufacturer, these sensors are operational at the temperature range between -20°C and 50°C.”

4. Equation 6, how to calculate the corresponding sensitivity at a specific temperature S_current?

Respond to Comment 4:

The following text was added to further explain this issue:

“For each sensor the manufacturer datasheets [75], [76] include a graph which describes the relation of sensor sensitivity according to the temperature of the environment. The calculation of the corresponding sensitivity at a specific temperature is done by extracting the slope of the graph at each temperature case.”

Reviewer 2 Report

This paper shows Air pollution monitoring via Wireless Sensor Networks: The investigation and correction of the aging behavior of electrochemical gaseous pollutant sensors. I have following suggestions: 1) The background to Wireless Sensor Networks should be clearly shown. What are the new features to your method?  You had better to give Advantages and disadvantages of existing methods, why or form which view you start your research point. 2) In the abstract part, you had better to show your method in step by step in order to show the core work you have done. Form the current abstract, I cannot find the innovation, which is superficial to tell your method. You had better to add the more step details. In introduction part, you have to reconstruct it according to the sequence of background, problem to state，existing methods with disadvantage，your method and from which point to solve. I find the related works are disorder. You can cite DOI:10.1145/3517154. DOI:10.1109/TNSE.2022.3163144. It should be sorted based on different method or methods. After you give and discuss proposed methods，you should state the problem or summary. 3）for core methods, how to find the anomaly? How to encode and decode the input and output data? 3) The experiment should be improved a lot. Please recheck experiment section that the experiment is hard to read. you have to reconstruct it according to the sequence of experiment target，Data preparation， how to compare？and your metrics. You should give the step of experiment that how to carry out, such as parameter，step, condition in scenario. You should show the configuration, that how to compute? You should reconstruct it according to the sequence of Phenomena, causes, and recommendations. You should give more details and Analysis according to experiment above you carry out.

Author Response

We would like to thank the reviewer for the time he/she spent to study our manuscript. His/Her comments and suggestions were carefully considered so that the answers given and the actions undertaken address all issues stated.  

This paper shows Air pollution monitoring via Wireless Sensor Networks: The investigation and correction of the aging behavior of electrochemical gaseous pollutant sensors. I have following suggestions:

1. The background to Wireless Sensor Networks should be clearly shown.

Respond to Comment 1

The background referring to WSNs was enriched by providing additional information in Section 1.

2. What are the new features to your method?

Respond to Comment 2

In order to clarify this issue, the following text was added:

“While the aging of electrochemical sensors is a given problem, the treatment of measurements during aging remains on the table. Based on this, the investigation of equations including correction factors to compensate and improve the measurements during their lifetime was the subject of research of this work”.

3. 3. You had better to give Advantages and disadvantages of existing methods,

Respond to Comment 3

The concept the reviewer rises is crucial and we gladly added the following text relatively with this consideration in the manuscript.

“Actually, most of the published works that deal with the aging of low cost sensors involve machine learning, neural networks and other similar processing demanding methods. Such approaches require significant CPU and memory resources having direct impact on the processing capability specifications and the cost of such a measuring system. Additionally, incorporating such methodologies on the measuring unit significantly increases the power consumption. Even for the case when the processing is conducted at a central point, the need of high processing power remains, since each networked measuring system must be treated separately. The herein proposed solution incorporates a simple compensation algorithm of good performance that can be easily executed at any sensing node. Also, due to the simplicity of the required actions power consumption is practically unaffected.”

 

4. Why or form which view you start your research point.

Respond to Comment 4

This research work was triggered by the fact that although air monitoring can indeed be performed by using WSNs containing low-cost gaseous pollutant electrochemical sensors, the accuracy of sensors of this kind is reduced during their lifetime because of aging.

5.  In the abstract part, you had better to show your method in step by step in order to show the core work you have done. Form the current abstract, I cannot find the innovation, which is superficial to tell your method. You had better to add the more step details.

• Respond to Comment 5
• Thank you for your valuable comment. In order to make Abstract be more representative of the research work that was carried out, the following text was added :

“The following steps were conducted in order to both study and lessen the aging of electrochemical sensors: i. A sensor network was developed to measure air quality at a place near official instruments that perform corresponding measurements. ii. The collected data were compared to the corresponding recordings of the official instruments. iii. Calibration and compensation was performed using the electrochemical sensor vendor instructions. iv. The divergence between the datasets was studied for various periods of time and the impact of aging was studied. v. The compensation process was re-evaluated and new compensation coefficients were produced for all periods. vi. The new compensation coefficients were used to shape formulae that automatically calculate the new coefficients with respect to the sensors’ aging. vii. The performance of the overall procedure was evinced through the comparison among the revised measurements and corresponding real data.”

 

Also, please refer to the last part of the Conclusions:

“   this research work evinced that there is a very effective methodology to keep the sensors’ performance stable during their lifecycle. Actually, coefficients C1 and C2 used in the methodology proposed, express the performance of a sensor during its operational life. Specifically, coefficient C1 is directly related to aging and its value changes for each month of the operational time of the sensor according to a formula which is differentiated according to the pollutant gas detected. At the same time, coefficient C2 aims at the micrometric correction of the sensor values according to the average temperature of the month of operation under study.

The suitable use of these two coefficients in the formulae proposed showed excellent results, for both NO2 and O3 low-cost air quality sensors in the sense that not only their aging is treatable but also high reliability of the measurements can be achieved for the entire lifetime of the sensors. In this way, air quality monitoring can be performed via low-cost sensors with no need for recalibration with official reference instruments at regular intervals. So, it is feasible to create dense air quality monitoring networks in urban areas, without high acquisition costs. This is greatly beneficial to the attainment of not only inexpensive but also accurate air monitoring via WSNs in Smart Cities.”

 

6. In introduction part, you have to reconstruct it according to the sequence of background, problem to state，existing methods with disadvantage，your method and from which point to solve. I find the related works are disorder. You can cite DOI:10.1145/3517154. DOI:10.1109/TNSE.2022.3163144. It should be sorted based on different method or methods. After you give and discuss proposed methods，you should state the problem or summary.

Respond to Comment 6

To address this issue the following text was added to the manuscript

 

“In addition, learning systems based of generative adversarial network (GAN) research team [63] proposed a GAN-based automatic property generation (GAPG) approach to generate verification properties supporting model checking. On the other hand the time-series feature of the IoT makes the data density and the data dimension higher, where anomaly detection is important to ensure hardware and software security, research work [64] proposed a memory-augmented autoencoder approach to detect anomalies in IoT data, which aims to use reconstruction errors to determine data anomalies”

 

7. for core methods, how to find the anomaly? How to encode and decode the input and output data?

Respond to Comment 7

 

The detailed description of the hardware used is presented in previous works of the Authors of this article, and are considered as padding information. The communication encoding and the data decoding are presented in our previous works cited in this manuscript [references 36-38 and 76]. Authors decided not to re-mention this information since it would not add any new information in current work.

8.  The experiment should be improved a lot. Please recheck experiment section that the experiment is hard to read. you have to reconstruct it according to the sequence of experiment target，Data preparation，how to compare？and your metrics. You should give the step of experiment that how to carry out, such as parameter，step, condition in scenario. You should show the configuration, that how to compute? You should reconstruct it according to the sequence of Phenomena, causes, and recommendations. You should give more details and Analysis according to experiment above you carry out.

 

Respond to Comment 8

Following your valuable recommendations the overall sections 3 and 4 were rewritten in order to describe the establishment and execution of the experimental procedures in a consistent and analytical way. Consequently, Section 5 was also improved.

Reviewer 3 Report

The article discuss aging behaviour of air pollution sensor networks operating in field. The authors propose correction strategy to mitigate the the aging issue by intermittent update of the calibration coefficient C1 and C2.

The work is systematically and well presented to a wide reader.

I would recommend this article to be publish with minor suggestions and questions:

1. Fig.2. How far is A and B? Is there potential discrepancy that may arise due to different pollution variation between those two sites?

2. Line 226: Can the author mention about the cross-sensitivity of the low-cost sensor? This is also a valuable information to assess the inaccuracy of the sensor.

3. Figure 7. The low cost sensor failed to reach the baseline periodically during the period of 27 April to 3 May.  Can it be attributed to the slow recovery of the low-cost sensor?

4. Table 2. The July data is peculiar, for instance the R2 suddently drop significantly e.g. in April it was 0.48 then 0.05 in July and increase again to 0.38 in Oct. Why?

I would also suggest the quantitative data like in Table 2 or 3 be presented in bar-plot since the huge volume of numbers are quite difficult to follow.

5. Inconsistency in Fig.13: in the panel it mentions April while in the caption it is October.

6. Fig.13 and 14: Please simplify and check the consistency of the significant figures

7. FIg.9 The C1 coefficient shows linear correlation overtime. Would the author expect this linear correlation for longer period? What would be the limitation of this linear model?

Author Response

We would like to thank the reviewer for the time he/she spent to study our manuscript. His/Her comments and suggestions were carefully considered so that the answers given and the actions undertaken address all issues stated.

1. Fig.2. How far is A and B? Is there potential discrepancy that may arise due to different pollution variation between those two sites?

Respond to Comment 1:

The distance between A and B is denoted. The validity of the measurements’ comparison is supported by relevant research works mentioned. Specifically, the following text was added to clarify this issue:

“The distance between these two locations is 900m. Both locations share the same urban conditions. Actually, the spatial coverage of a monitoring site represents the quantification of the variability of concentrations of a specific pollutant around the site [69], [70]. while the assessment of representativeness aims at the delimitation of areas of the concentration field with similar characteristics at specific locations as well the spatial surrogate data (similar emissions sources and land-cover characteristics) [71], [72], [73].”

 

2. Line 226: Can the author mention about the cross-sensitivity of the low-cost sensor? This is also a valuable information to assess the inaccuracy of the sensor.

Respond to Comment 2:

In order to clarify this issue the following text was added:

 

The cross-sensitivity from other gasses can be seen at the technical specifications of the manufacturer datasheet, of the ozone sensor (OX-B431) [75] and of the nitrogen dioxide sensor (NO2-B43F) [76]. Specifically for the NO₂ sensor (NO2-B43F) presented cross-sensitivity, (% measured gas @5 ppm) for the gases H₂S<-80, NO<5, CI₂<100, SO₂<-3 and CO<-3, (% measured gas @100 ppm) for the gases H₂<0.1, C₂H₄<0.1, NH₃<0.5, and the Halothane not detected. The O₃ sensor (OX-B431) presented cross-sensitivity, (% measured gas @5 ppm) for the gases H₂S<-80, NO<5, CI₂<100, SO₂<-3 and CO<-3, (% measured gas @100 ppm) for the gases H₂<0.1, C₂H₄<0.1, NH₃<0.5, and the Halothane <0.1.

3. Figure 7. The low cost sensor failed to reach the baseline periodically during the period of 27 April to 3 May. Can it be attributed to the slow recovery of the low-cost sensor?

 

Respond to Comment 3:

Indeed the sensors after the compensation and calibration processes show inclinations in the cases of O₃ low concentrations. The main reason for these inclinations is the underlying physical mechanisms that activate the sensors. It is indicative that the cross sensitivity of the used sensors involve other oxide pollutants that exist at the environment but are of low concentrations. Such a behavior has been verified in several cases. Under the above concept this behavior can also be attributed to the localized cases (point A and B) and the distance between the reference and low cost stations. In any case the trend-line of the recordings is consistent. /

In order to resolve this issue the following text was added:

“Observing Figure 7 it becomes evident that the low cost sensor fails to reach the minimum values when these are obtained from the reference instruments. This can be attributed to the cross sensitivity of the sensor that is activated from other existing environmental oxides.”

 

4. Table 2. The July data is peculiar, for instance the R2 suddently drop significantly e.g. in April it was 0.48 then 0.05 in July and increase again to 0.38 in Oct. Why?

Respond to Comment 4:

The reviewer has spotted the issue that activated the Authors to study the field of aging. An important issue that is not discussed sufficiently in the literature is the impact of the season on the sensors accuracy. Such a field is still under investigation in the literature. The very low NO₂ R² value during July remains a field of study but it is not affecting current work since as can be seen it is not deteriorating or improving after applying the aging compensation equations.  It is important to notice that even after the compensation and correction of the factors the R² is not becoming significantly higher for this case. The temperature limitations of the sensor vendors are also an issue that affects this performance.

In order to clarify this issue the following text was added:

“It becomes evident from Table 2 that NO₂ values during July show low cross correlation. As it will be discussed later on the impact of the aging compensation equations on the NO₂ is positive as it improves the corresponding R² despite the fact that it remains at low values.  Furthermore, in Greece, during the summertime, the environmental temperature push the sensors at their functional limits. Specifically, according to the manufacturer, these sensors are operational at the temperature range between -20°C and 50°C.”

5. I would also suggest the quantitative data like in Table 2 or 3 be presented in bar-plot since the huge volume of numbers are quite difficult to follow.

Respond to Comment 5:

Authors would like to thank the reviewer for his comment. Initially, it was considered that the high number of Figures would affect the review process. In any case the Authors agree with the reviewer and cordially added Figures 11 and 12 in the manuscript.

 

 

6. Inconsistency in Fig.13: in the panel it mentions April while in the caption it is October.

Respond to Comment 6:

Thank you for your comment. The Authors corrected the Figure caption.

 

 

7. Fig.13 and 14: Please simplify and check the consistency of the significant figures

Respond to Comment 7:

Thank you for your comment. The Authors simplified the Figures providing only the essential information removing the text from the figure and leaving it only on the axis and marked the median with an X.

 

 

8. FIg.9 The C1 coefficient shows linear correlation overtime. Would the author expect this linear correlation for longer period? What would be the limitation of this linear model?

Respond to Comment 8:

Thank you for your valuable comment. Indeed there is a time limitation of the proposed model. But it must be taken into consideration that the manufacturer gives a two year maximum lifetime of the sensors. Since this can rise as a noticeable question for the future readers of the manuscript the following text was added in the manuscript. 

 “It is observed that C1 coefficient shows linear correlation. The expected temporal limitation of the C1 trend is significantly longer than the two year lifetime of the sensors as provided by the manufacturers. “

 

Round 2

Reviewer 2 Report

I am happy to accept this paper.